How to OCR PDFs for Free

How to OCR PDFs on Windows (for free)

I recently needed to run OCR on a 538-page scanned Cyrillic PDF. No budget, no subscriptions, just a Windows machine and some patience. Here's what worked.

The tool stack:

The winning combination is Tesseract (open-source OCR engine) wrapped by OCRmyPDF (a Python tool that handles the full pipeline: takes a scanned PDF in, runs Tesseract page by page, and spits out a searchable PDF). Together they handle deskewing, preprocessing, and text layer embedding without you needing to split pages into images manually.
There are fancier options. ABBYY FineReader is the commercial gold standard for Cyrillic and will beat Tesseract on accuracy, especially at low DPI. Google Cloud Vision has a free tier of 1,000 pages/month and handles degraded scans better than anything else. But if you want something fully local and free with no page limits, Tesseract + OCRmyPDF is the move.

Setup

1. Install Python

Download from python.org/downloads. During installation, check "Add python.exe to PATH" at the bottom. This is the step everyone skips and then spends an hour debugging. Don't skip it.
Open a fresh Command Prompt and verify:
python --version
pip --version

2. Install Tesseract

Download the Windows installer from github.com/UB-Mannheim/tesseract/wiki.
During installation, expand the language data options and check Serbian (Cyrillic) (srp). If you also need French, Latin, Albanian, or other languages, check those too. You can always add languages later by downloading .traineddata files from github.com/tesseract-ocr/tessdata and dropping them into C:\Program Files\Tesseract-OCR\tessdata\.
Optional but recommended: add C:\Program Files\Tesseract-OCR to your system PATH. Search "Environment Variables" in Windows, find Path under System variables, click Edit, click New, paste the path.

3. Install OCRmyPDF

pip install ocrmypdf

Run the OCR
Navigate to the folder containing your PDF:
cd %USERPROFILE%\Downloads
Then run:
ocrmypdf --language srp --deskew input.pdf output.pdf
That's it. The --deskew flag straightens crooked scans.  Regarding the language selection just type tesseract --list-langs.

Hopefully this helps anyone.

Comments

Post a Comment

Popular posts from this blog

Welcome to My Blog - Introduction

As I begin to first fiddle around with this blog, I might as well introduce myself to you guys some more! My name is Dorart, and I'm a junior in Pace University. I'm majoring in Information Technology at the Pleasantville campus location. I was born an raised in New York my whole life, specifically in the Bronx. I know New York as home, but I have since left the city and moved up towards campus in Westchester County.  In terms of my background, I've always had an interest in technology, specifically how to tinker with hardware. At a young age I found myself fixing all sorts of problems from storage upgrades on laptops to building my own whole gaming rig. I've used social media at a pretty early age, as it particularly tied to my interest in gaming & hardware. I found Skype to be the ultimate platform to communicate with my friends & family back then, and also the infamous standard of social media: YouTube. I also occasionally edited Wikipedia during my high scho...
Read more